Citation Context

...ompression of the graph as it is written, in much the way image file formats allow intrinsic compression of the image. Graph Compression algorithms have been a topic of study at least since 2001 [13]–=-=[15]-=-, with numerous follow-ups. So it is interesting that only one format is designed around this feature. However, two other formats provided some crude mechanisms to reduce the size of the file. Finally...

Citation Context

... and queried. On the other hand, the standard assumption in data exchange is that the data itself is relatively static, but portability is important. There is a wide-ranging survey of graph databases =-=[1]-=-, which is more concerned with the underlying database aspects, e.g., the relationship between a graph database and other more traditional relational database, and the properties of various exemplar g...

Citation Context

...for compression of the graph as it is written, in much the way image file formats allow intrinsic compression of the image. Graph Compression algorithms have been a topic of study at least since 2001 =-=[13]-=-–[15], with numerous follow-ups. So it is interesting that only one format is designed around this feature. However, two other formats provided some crude mechanisms to reduce the size of the file. Fi...

Citation Context

...aphs, other representations are possible (for instance tensors can generalise the concept of an adjacency matrix for multi-layer networks). However, codification of these is an ongoing research topic =-=[12]-=- and so we will not try to encapsulate it here. C. File structure This field describes how the file format’s structure is defined. The cases are: simple : the typical approach to create a graph format...

Citation Context

...n these formats. B. Representation The second critical decision in storing a graph is the choice of representation: matrix : The graph’s full adjacency matrix. edge list : A list of the graph’s edges =-=[9]-=-. smatrix : The matrix representation is poor for sparse graphs, which are common in real situations. However, some tools actually store a sparse matrix, which is almost equivalent to an edge list. Th...

Citation Context

... general terms, but only 14 in detail, and with a different focus: for instance they provide more information on software support for formats than we provide here. One additional paper to consider is =-=[7]-=-, which was written specifically with the view of designing a new, more universal graph format. We deliberately avoid this approach in order to avoid bias in our discussion. III. THE FILE FORMATS As n...

Citation Context

...useful, for instance, with a tree or ring. constructive : Graphs can often be described in terms of mathematical operations used to construct the graphs: for instance graph products on smaller graphs =-=[10]-=-. See [11] for a description of “levels” of graph formats. procedural : Many graphs can be concisely defined by a set of procedures, rather than explicit definition of the nodes and links. This type o...

Citation Context

...r instance, with a tree or ring. constructive : Graphs can often be described in terms of mathematical operations used to construct the graphs: for instance graph products on smaller graphs [10]. See =-=[11]-=- for a description of “levels” of graph formats. procedural : Many graphs can be concisely defined by a set of procedures, rather than explicit definition of the nodes and links. This type of graph fo...

Citation Context

...say that these issues have taken second place in the design of graph exchange formats. There have been a number of other efforts to gather similar information on graph exchange formats by researchers =-=[2]-=-–[4] and software distributors [5], [6]. The results provided inspiration for some of the descriptors used here, but this work aims at being more comprehensive. We surveyed approximately 100 possible ...

Citation Context

...inspiration for some of the descriptors used here, but this work aims at being more comprehensive. We surveyed approximately 100 possible formats, and include detailed data on 76, as compared to 7 in =-=[3]-=-, 28 in [4], 5 in [5] and 11 in [6]. Bodlaj [2] considers a similar number of formats in general terms, but only 14 in detail, and with a different focus: for instance they provide more information on...

Citation Context

...that these issues have taken second place in the design of graph exchange formats. There have been a number of other efforts to gather similar information on graph exchange formats by researchers [2]–=-=[4]-=- and software distributors [5], [6]. The results provided inspiration for some of the descriptors used here, but this work aims at being more comprehensive. We surveyed approximately 100 possible form...

Citation Context

...ite this focus, there were still 76 file formats considered. Consequently, we do not have space to present details of all of the file formats here. The details, and original data can be obtained from =-=[8]-=-. This paper provides summary statistics and analysis of the file formats. We sought feedback on our data from authors or maintainers of the file formats, and received input for 23 of the formats. Thi...

Citation Context

...nodes that themselves contain graphs. Several formats provide mechanisms to record this substructure. Unfortunately, there does not seem to be a consistently used definition of this type of structure =-=[17]-=-, and so we see differences not just in the representation, but also what exactly is being represented. The problem becomes even more complicated when hierarchy and hyper-graphs are combined [12] (the...

Citation Context

...ic databases that provide example networks for benchmarking or research. We provide a list in [8] of some of the better known of these with their format choices. Additional data sources are listed in =-=[18]-=-, and a detailed taxonomy and examples of computer-network data appears in [19]. There is no clear winner here: slightly preferred is a variant of the Trivial Graph Format due to its least-commondenom...

Citation Context

...vide a list in [8] of some of the better known of these with their format choices. Additional data sources are listed in [18], and a detailed taxonomy and examples of computer-network data appears in =-=[19]-=-. There is no clear winner here: slightly preferred is a variant of the Trivial Graph Format due to its least-commondenominator status (but note that this isn’t really one format, so much as a collect...